NeurIPS Topic Selection
Use this skill before committing to NeurIPS. The target is not "any good AI paper"; it is a paper
whose contribution will matter to the NeurIPS reviewer community and survive the current official
track rules.
Fit signals
Strong NeurIPS candidates usually have one of these cores:
- a general ML method, model, objective, optimization, inference, or learning principle;
- a theory result that changes understanding of ML behavior or limits;
- a high-quality empirical finding about models, data, evaluation, robustness, or scaling;
- a use-inspired result with a real scientific, social, health, robotics, sustainability, or
creative-AI problem and a clear ML contribution;
- a dataset, benchmark, or evaluation contribution that belongs in the correct current NeurIPS
track rather than being forced into main track;
- a rigorous negative result that changes community understanding.
Poor fit signals
- Engineering integration without a research insight.
- Domain application where the ML contribution is ordinary.
- Benchmark improvement without mechanism, error analysis, or generality.
- Safety, fairness, or societal claim with thin evidence.
- Reproduction or replication study better suited to MLRC/TMLR.
- Dataset or evaluation paper that should use the E&D track.
- Position argument that should use the Position Papers track.
Contribution-type choice
Choose the contribution type that changes reviewer expectations. A theory paper should make proofs
central. A use-inspired paper needs a real task and ML novelty. A concept-and-feasibility paper
needs high-risk/high-reward framing and credible preliminary evidence. A negative-results paper
needs a lesson that matters beyond one failed run.
Output format
[Fit] High / Medium / Low
[Recommended track] Main / E&D / Position / MLRC / workshop / other venue
[Contribution type] General / Theory / Use-Inspired / Concept & Feasibility / Negative Results
[Why NeurIPS] <one sentence>
[Main upgrade needed] <evidence, framing, related work, artifact, ethics, or reroute>
Source: brycewang-stanford/Awesome-Journal-Skills → NeurIPS-Skills/skills/neurips-topic-selection/SKILL.md
1---2name: neurips-topic-selection3description: Use when deciding whether a paper belongs at NeurIPS, choosing main-track versus another NeurIPS track, selecting contribution type, or rerouting to a better AI/ML venue.4---567# NeurIPS Topic Selection89Use this skill before committing to NeurIPS. The target is not "any good AI paper"; it is a paper10whose contribution will matter to the NeurIPS reviewer community and survive the current official11track rules.1213## Fit signals1415Strong NeurIPS candidates usually have one of these cores:1617- a general ML method, model, objective, optimization, inference, or learning principle;18- a theory result that changes understanding of ML behavior or limits;19- a high-quality empirical finding about models, data, evaluation, robustness, or scaling;20- a use-inspired result with a real scientific, social, health, robotics, sustainability, or21 creative-AI problem and a clear ML contribution;22- a dataset, benchmark, or evaluation contribution that belongs in the correct current NeurIPS23 track rather than being forced into main track;24- a rigorous negative result that changes community understanding.2526## Poor fit signals2728- Engineering integration without a research insight.29- Domain application where the ML contribution is ordinary.30- Benchmark improvement without mechanism, error analysis, or generality.31- Safety, fairness, or societal claim with thin evidence.32- Reproduction or replication study better suited to MLRC/TMLR.33- Dataset or evaluation paper that should use the E&D track.34- Position argument that should use the Position Papers track.3536## Contribution-type choice3738Choose the contribution type that changes reviewer expectations. A theory paper should make proofs39central. A use-inspired paper needs a real task and ML novelty. A concept-and-feasibility paper40needs high-risk/high-reward framing and credible preliminary evidence. A negative-results paper41needs a lesson that matters beyond one failed run.4243## Output format4445```text46[Fit] High / Medium / Low47[Recommended track] Main / E&D / Position / MLRC / workshop / other venue48[Contribution type] General / Theory / Use-Inspired / Concept & Feasibility / Negative Results49[Why NeurIPS] <one sentence>50[Main upgrade needed] <evidence, framing, related work, artifact, ethics, or reroute>51```5253---5455**Source:** [`brycewang-stanford/Awesome-Journal-Skills`](https://github.com/brycewang-stanford/Awesome-Journal-Skills) → `NeurIPS-Skills/skills/neurips-topic-selection/SKILL.md`
thedixitjain